The Draft,The Draft,The Draft………

Fractals are created by the repetition of geometric patterns via a process of iteration. Because they look similar at any zoom level, fractals are often considered to be infinitely complex. Multiple objects in nature are approximated by fractals (clouds, mountain ranges, lightning bolts, Troy Polamanu’s hair).

So it’s no surprise that some of our analysis of the draft is starting to follow patterns found in nature. My last post was a  draft post in response to a reader comment and this post? A draft post in response to a reader comment. So hey, aren’t  fractals awesome?

The comment in question is from neal frazier:

so only 9 of 35 top 5 picks actually became top 5 players(hereinafter TT5) – an awful success rate of ~25%, but only 15 of 175 picks 6-30 became TT5 – a success rate of ~8% and only 11 of 210 second round draft picks became TT5 – a success rate of ~5%. If getting one of the TT5 is your goal, then a top 5 pick is still about 3 times as valuable as a later first round pick.

This begs the question of how big is the difference between someone who became TT5 from someone who became TT6-30 – ie what is the value of a near miss? The 6-30 picks should have a virtually identical, if not somewhat higher, chance of landing someone who was actually a TT6-30 pick than a top 5 pick has. If a TT6-30 pick has nearly the same productivity as a TT5 pick, then the value of a 6-30 pick moves closer to a top 5 pick – just eyeballing your chart it looks like the TT5 guys average around 0.170 WP48. This is substantially higher than the average 0.100 WP48 that TT6-30 guys are likely to be around. Since you would always give up two 0.100 WP48 guys for a single 0.150 WP48 guy, this near miss value is minimal.

I guess what I am saying is that these numbers convince me that I would trade 2 later first round picks for a top 5 pick in a heartbeat – it gives me 1.5 times (25% to 16%) the chance of getting a TT5 guy and the combined salary of the 2 guaranteed contracts I am getting out of is probably higher than the salary of the one guy that I end up with (not sure how steep the salary curve is…).

related conclusion – I would never give up 2 second round picks for a later first round pick because I would be decreasing my chances (~10% to ~8%) of getting a TT5 guy and taking on more guaranteed salary – a double loss.

Seperate topic: I wonder if GMs used to have more success identifying TT5 guys back before players started leaving early – were the top 5 picks from the 80′s TT5 guys more frequently than they are now that the top 5 picks consist almost entirely of freshmen and sophomores?

Awesome comment. Let’s take it by parts. First the background.

Some quick background
This article uses Wins Produced and WP48 [Wins Produced per 48 minutes] to evaluate player’s performance.* This measure uses three key components to evaluate a player:

  • The player’s per minute box score statistics
  • The player’s team’s per minute box score statistics
  • The average performance at the player’s position (PG, SG, SF, PF or C)

A full explanation can be found here. To give a general scale, an average player has a WP48 score of 0.100. The very best players in the league usually have a WP48 over 0.300. To put this in perspective; an average player who plays a full season at 24 minutes a game would generate around four wins for their team.  In contrast, a player posting a 0.300 WP48 would generate more than twelve wins in this time on the court.

Average Value of Players Drafted in WP48

If we look at the the best twenty players available for each draft (1st 4 Years in the League
Greater than 1599 Minutes Played ranked by WP48) we get the following table:

And as a chart:

So netting a TT5 (Top 5 Player) in the last 2 decades gets you a >.150 WP48 player (6 wins at 24 minutes played per game or about $10.2 million in value at $1.7 mill per win).  So netting a TT12 (Top 12 Player) in the last decade gets you a >.100 WP48 player (4 wins at 24 minutes played per game or about $6.8 million in value at $1.7 mill per win). After that you get marginal starters and bench guys (assuming of course that no one slipped through the tracks, a large assumption actually). The best player is about twice as valuable on average than number 8 and 2 twice the value of ten. So any two picks that net two player in the top 10 are netting you more total wins (granted all things being equal you’d rather have one player that wins 12 than two who win six each). Since, I wouldn’t however give up two .100 for a  .150  getting a top five guy is only about 30 to 40% more valuable than getting a 6 to 10 guy (in terms of wins).

Average Pick of the Top Twelve Players Drafted ranked by WP48

If we now look at the the best twelve players drafted (1st 4 Years in the League
Greater than 1599 Minutes Played ranked by WP48) in terms of draft position we get the following table:

By the numbers, Pick 10 gives you a chance at a top three guy every year.  The majority of the time, Quality players (TT5) are always available in the draft no matter where you pick in the first round and availability is the key here. If I’m picking randomly then volume matters but if I can figure out how to pick better than the other guy? Then a latter position where I know my guys will be available for cheap is preferable.

My conclusion is that:

  • Talent is always available in the first round.
  • The trick is finding it with some accuracy (which I postulate we can do).
  • Given that the identification is risky a later,cheaper pick is better.

Finally, the last topic was if GMs used to have more success identifying TT5 guys. One final table reveals:

Not really.

Tomorrow we break out of our recursive loop and go back to our regularly scheduled topics (I have no clue which one though)

Written by:

85 Comments

  1. Austin
    9/20/2010
    Reply

    I think you’re dealing with some confirmation bias here. If a late first round or 2nd round guy performs poorly, he probably won’t get 1600 minutes. You’re looking at the players who played well enough to get that much time in the league. At the very least you need to stipulate something like “60% of #15 picks played 1600 minutes in their first four years”.

    It is a fairly low threshold, especially for lottery picks, but I’d imagine that enough >16 picks flame out to affect the data.

    Alternately it might be worth it to look at the standard deviation of the production of draft picks, or even better, the distribution of the production at those draft picks. Some of the later picks could heavily skew the sample, especially when you’re looking at only 10 data points – a Manu Ginobili or someone like that.

    Finally, since you’d be using PAWS40 or something similar to evaluate college players at least in part, what kind of results would you get in actual NBA production compared to the actual picks made? I think this is crucial, because if you’re able to better predict stars, then earlier picks could become more valuable again – as a guarantee you could get a productive player (or even have your choice, so that having to select a player at a position you already have covered isn’t a problem) rather than just hoping that one drops to you at 15 or 30.

    I’m going after this so hard because to me, as I suspect to many other people, it does not make intuitive sense to give up high draft picks so easily. Trade one lottery pick for 2 or 3 late first rounders, sure (if you have roster space) – but your position is extreme, and so I’m going after the methodology. Fascinating topic for sure.

  2. Austin
    9/20/2010
    Reply

    I think you’re dealing with some confirmation bias here. If a late first round or 2nd round guy performs poorly, he probably won’t get 1600 minutes. You’re looking at the players who played well enough to get that much time in the league. At the very least you need to stipulate something like “60% of #15 picks played 1600 minutes in their first four years”.

    It is a fairly low threshold, especially for lottery picks, but I’d imagine that enough >16 picks flame out to affect the data.

    Alternately it might be worth it to look at the standard deviation of the production of draft picks, or even better, the distribution of the production at those draft picks. Some of the later picks could heavily skew the sample, especially when you’re looking at only 10 data points – a Manu Ginobili or someone like that.

    Finally, since you’d be using PAWS40 or something similar to evaluate college players at least in part, what kind of results would you get in actual NBA production compared to the actual picks made? I think this is crucial, because if you’re able to better predict stars, then earlier picks could become more valuable again – as a guarantee you could get a productive player (or even have your choice, so that having to select a player at a position you already have covered isn’t a problem) rather than just hoping that one drops to you at 15 or 30.

    I’m going after this so hard because to me, as I suspect to many other people, it does not make intuitive sense to give up high draft picks so easily. Trade one lottery pick for 2 or 3 late first rounders, sure (if you have roster space) – but your position is extreme, and so I’m going after the methodology. Fascinating topic for sure.

  3. Chicago Tim
    9/20/2010
    Reply

    Hey, Arturo. This is kind of off topic, but I wondered if you had seen this article about Miami Heat coach Erik Spoelstra thinking about playing James at point and Bosh at center:

    “‘LeBron certainly will play minutes during the game at point guard and handle the ball a lot,’ Spoelstra said. ‘He will be a playmaker. But he has also been a scoring champ. He will be at the end of plays to finish them. He will be a facilitator. He’ll be so many different things. We want to take advantage of all his skills.'”

    “Spoelstra will use preseason (camp opens Sept. 28) to decide whether to start Mario Chalmers or Carlos Arroyo or an intriguing lineup with swingman Mike Miller starting and James and Wade handling the ball.”

    “Starting power forward Chris Bosh ‘won’t play the majority of his minutes at [center]’ but will play some there, Spoelstra said. He and Udonis Haslem ‘could be a very good defensive and rebounding duo.”’

    Spoelstra apparently would like to get Miller and Haslem on the floor. But he seems hesitant to start James at point and Bosh at center, perhaps because he knows those aren’t the positions they want to play.

    http://www.miamiherald.com/2010/09/18/v-fullstory/1831335/spoelstra-expects-to-use-lebron.html

      • Chicago Tim
        9/21/2010
        Reply

        You may want to hold off until Friday. Dampier is supposed to choose his new team by Thursday. It’s between Houston, Miami, Atlanta and Denver.

        • Gray Jay
          9/21/2010
          Reply

          Unless Dampier’s a graduate of the Antoine Walker School of Money Management, I have a feeling Dampier’s decision will have roughly as much suspense and surprise as Pat Riley’s did after “the Decision.” Dampier’s made damn near 100 million playing basketball and has no ring. Is he really going to be swayed by the extra 1M a year the Rockets or Atlanta can pay him? And this isn’t counting any shoe deal chicanery or the valid point that if he goes to Miami he will have a quantum leap in the national exposure he can expect over playing for any other team.

          If Dampier and his ~.150 WP48 over the last several years goes to Miami, he’s now their starting center, if Pat Riley has any brain cells not terminally poisoned by years of Vitalis saturation. Not Zombie Ilgauskas (-.040), not Joel Anthony (-.017—but hey, he’s a rook, maybe it’ll get better), certainly not Bosh (Bosh didn’t help form the “Cheats” so he’d have to do dirty work in the low block): Dampier will.

          And when he does, he will not be required to do anything other than play D and hoover boards like a Roomba on nitrous. He can then, assuming no giant lock-out, parlay that exposure into one more large veteran multi-year deal elsewhere. You know there’s an NBA GM who’ll give him one. See, e.g. Gortat, Marcin.

          As for the original comment, I thought the point of the prototypical 1 was to help everyone else get a high-percentage shot, typically close to the basket. 4 of the 5 guys on the floor for Miami don’t need any help getting those kinds of shots. James and Wade can typically get them all by themselves. For as much as I dislike Bosh and think he’s soft, I concede he is capable of destroying most 1v1 defenders in the post. Miller is just devastating anywhere on the perimeter, provided he has a halfway open look. With James and Wade driving the lane at will, Miller will probably be left open. A lot. All of this is to argue that the Heat really don’t need a set 1: just let either James or Wade walk the ball up the floor, in essence having Wade and James trade off between the 1 and 2. Miller will float around at the 3, Bosh will play the 4, and Haslem/Ilgauskas/Anthony/Dampier will serve as the inexhaustible supply of big men. Chalmers/Arroyo are really there for garbage time—which there should be a lot of—and occasionally spelling Wade/James.

          The Heat this year are really going to be unfair and anti-competitive. It’s one of the NBA characteristics that are increasingly driving my viewing towards the EPL.

            • Gray Jay
              9/22/2010
              Reply

              Jumping the gun a bit, but do your calculations, Arturo, come out to an Estimated Heat wins of around 80? This is without Dampier, by the way. With Dampier, I get about 86 wins. (Negative production from Anthony and Ilgauskas goes to him instead) All I did was set up a minutes allocation for the top 10 Heat players, using historical minutes averages, and then multiply by the WP48 figures in Alvarez’s database.

              I’m sure I must be screwing up the WP48 positional adjustments. Is there an easy conversion factor to convert WP48 for time at the 5, to an estimated WP48 for time at the 4, and so on?

              Considering the single season wins record is 72, by the 95-96 Bulls, has anyone gone through the 95-96 Bulls, calculated the Estimated Wins, and compared them to this Heat squad? I’m just curious how much faith to put into Berri’s model at the extreme high end.

              • Gray Jay
                9/22/2010

                Oh, and thanks for the compliment. Didn’t mean to rant as much as I did.

              • 9/22/2010

                For team predictions, You should always use ADJP48 times the minutes and do the position adjustment as a team (Sum of ADJP48* Minutes played/48 – (PosAdjust for PG*82+PosAdjust for SG*82+…..). Just use the position adjust numbers for last year.

                I’m doing a team review on the Heat but I’m waiting for the Dampier thing to get resolved first.

  4. Chicago Tim
    9/20/2010
    Reply

    Hey, Arturo. This is kind of off topic, but I wondered if you had seen this article about Miami Heat coach Erik Spoelstra thinking about playing James at point and Bosh at center:

    “‘LeBron certainly will play minutes during the game at point guard and handle the ball a lot,’ Spoelstra said. ‘He will be a playmaker. But he has also been a scoring champ. He will be at the end of plays to finish them. He will be a facilitator. He’ll be so many different things. We want to take advantage of all his skills.'”

    “Spoelstra will use preseason (camp opens Sept. 28) to decide whether to start Mario Chalmers or Carlos Arroyo or an intriguing lineup with swingman Mike Miller starting and James and Wade handling the ball.”

    “Starting power forward Chris Bosh ‘won’t play the majority of his minutes at [center]’ but will play some there, Spoelstra said. He and Udonis Haslem ‘could be a very good defensive and rebounding duo.”’

    Spoelstra apparently would like to get Miller and Haslem on the floor. But he seems hesitant to start James at point and Bosh at center, perhaps because he knows those aren’t the positions they want to play.

    http://www.miamiherald.com/2010/09/18/v-fullstory/1831335/spoelstra-expects-to-use-lebron.html

      • Chicago Tim
        9/21/2010
        Reply

        You may want to hold off until Friday. Dampier is supposed to choose his new team by Thursday. It’s between Houston, Miami, Atlanta and Denver.

        • Gray Jay
          9/21/2010
          Reply

          Unless Dampier’s a graduate of the Antoine Walker School of Money Management, I have a feeling Dampier’s decision will have roughly as much suspense and surprise as Pat Riley’s did after “the Decision.” Dampier’s made damn near 100 million playing basketball and has no ring. Is he really going to be swayed by the extra 1M a year the Rockets or Atlanta can pay him? And this isn’t counting any shoe deal chicanery or the valid point that if he goes to Miami he will have a quantum leap in the national exposure he can expect over playing for any other team.

          If Dampier and his ~.150 WP48 over the last several years goes to Miami, he’s now their starting center, if Pat Riley has any brain cells not terminally poisoned by years of Vitalis saturation. Not Zombie Ilgauskas (-.040), not Joel Anthony (-.017—but hey, he’s a rook, maybe it’ll get better), certainly not Bosh (Bosh didn’t help form the “Cheats” so he’d have to do dirty work in the low block): Dampier will.

          And when he does, he will not be required to do anything other than play D and hoover boards like a Roomba on nitrous. He can then, assuming no giant lock-out, parlay that exposure into one more large veteran multi-year deal elsewhere. You know there’s an NBA GM who’ll give him one. See, e.g. Gortat, Marcin.

          As for the original comment, I thought the point of the prototypical 1 was to help everyone else get a high-percentage shot, typically close to the basket. 4 of the 5 guys on the floor for Miami don’t need any help getting those kinds of shots. James and Wade can typically get them all by themselves. For as much as I dislike Bosh and think he’s soft, I concede he is capable of destroying most 1v1 defenders in the post. Miller is just devastating anywhere on the perimeter, provided he has a halfway open look. With James and Wade driving the lane at will, Miller will probably be left open. A lot. All of this is to argue that the Heat really don’t need a set 1: just let either James or Wade walk the ball up the floor, in essence having Wade and James trade off between the 1 and 2. Miller will float around at the 3, Bosh will play the 4, and Haslem/Ilgauskas/Anthony/Dampier will serve as the inexhaustible supply of big men. Chalmers/Arroyo are really there for garbage time—which there should be a lot of—and occasionally spelling Wade/James.

          The Heat this year are really going to be unfair and anti-competitive. It’s one of the NBA characteristics that are increasingly driving my viewing towards the EPL.

            • Gray Jay
              9/22/2010
              Reply

              Jumping the gun a bit, but do your calculations, Arturo, come out to an Estimated Heat wins of around 80? This is without Dampier, by the way. With Dampier, I get about 86 wins. (Negative production from Anthony and Ilgauskas goes to him instead) All I did was set up a minutes allocation for the top 10 Heat players, using historical minutes averages, and then multiply by the WP48 figures in Alvarez’s database.

              I’m sure I must be screwing up the WP48 positional adjustments. Is there an easy conversion factor to convert WP48 for time at the 5, to an estimated WP48 for time at the 4, and so on?

              Considering the single season wins record is 72, by the 95-96 Bulls, has anyone gone through the 95-96 Bulls, calculated the Estimated Wins, and compared them to this Heat squad? I’m just curious how much faith to put into Berri’s model at the extreme high end.

              • Gray Jay
                9/22/2010

                Oh, and thanks for the compliment. Didn’t mean to rant as much as I did.

              • 9/22/2010

                For team predictions, You should always use ADJP48 times the minutes and do the position adjustment as a team (Sum of ADJP48* Minutes played/48 – (PosAdjust for PG*82+PosAdjust for SG*82+…..). Just use the position adjust numbers for last year.

                I’m doing a team review on the Heat but I’m waiting for the Dampier thing to get resolved first.

  5. Xavier Q
    9/20/2010
    Reply

    Austin,

    The issue doesn’t seem to be to get the best talent. If you have a #1 pick then 100% of TT5 guys are available, so it’s the best pick to have. The issue is value, since #1 picks get paid substantially more than #10 picks.

    There is somewhat of a performance bias, in that later picks don’t get playing time if they don’t perform, whereas higher picks get playing time no matter what. But that only comfirms that the lottery picks are playing poorly, while also eliminating that 30th pick that got 2 minutes of playing time and scored 2 points off an offensive rebound.

    Just looking at the numbers it is intuitive that GM’s do not know how to select talent. In 3 of the 7 sample years the #1 pick wasn’t in the top 12 performing players over their rookie contract. So if it is in essence simply chance whether you land a top talent or not, it would be preferable to select later in the round and pay that pick less money. And if you have a fairly accurate measure to predict which players are going to be productive, like PAWS48, then those later picks are even more valuable since the people picking ahead of you are likely to pass over at least one or two of the top players. Again, so you can pay the player less and get more value for your money.

    • 9/20/2010
      Reply

      Austin,
      You’ll note that I focused on per minute performance and took out players with small samples to even out the noise. I’m actually of the opinion that a lot of decent players are slipping through the cracks in the current system but that’s not the central issue. My conclusion is about market inefficiencies at identifying talent and how that drives TT5 player availability to the later part of the first.
      XQ,
      You nailed it .
      While that market inefficency exists, lower draft picks will always have more value given that there’s a huge drop-off in financial commitment and public expectation (my risk) versus no appreciable drop-off in potential value. If I can put a model together that has a better (say a 50% percent) chance of id’ing TT5 talent? Then the goal would be to identify undervalued talent and draft it as late as possible to maximize my possible value by mitigating my risk (salary and expectations) while possibly also getting value in return.

      Let’s give an example from real life. The Twolves supposedly coveted Evan Turner. If I’m the GM for the 76ers and KHAAAANNN calls me with an offer of the 4, 16 and 30 for the number 2, I cover up the phone to stop from laughing and take that to the bank (in my mind that’s a Turner for Cousins, Damion Jones and Whiteside trade). A 2 for 4 and 16 would also be a no brainer ( and here I could probably get the Twolves to throw in some sort of future pick or cash compensations).

      • StLreflections
        9/22/2010
        Reply

        The opportunity costs of the hype plus extra expense of a number 1 pick are negligible in comparison to the value of high performance. Rookie contracts are all vastly below market value for star players. As an NBA GM I’d happily trade an increased chance at a superstar in exchange for some bad publicity after a bust.

        Also, of course, limiting things to the 1st 4 years underestimates the value of some stars. e.g., Josh Howard is much less valuable than Bosh, LeBron, and Wade today. Even if you only have absolute control for the first (actually 5 years) of a contract, drafting a player has real advantages in terms of controlling the resource, both increasing the possibility of keeping that player around, and also giving you the ability to trade them for other valuable assets.

        I’m looking forward to seeing your method for picking top 5 guys/draft evaluation-if you can hit 50% on the late first round/second round breakouts we’ve seen over the last 15 years (that is, both correctly identifying who is undervalued, and not ending up with lots of false positives), that’s one thing-then I do think you can draft around that strategy, but if your odds don’t change things that much, then its not as effective. The raw numbers matter a lot in evaluating your method. Right now, its obvious that #1 draft picks are tremendously valuable, since they are so much more likely to provide high production in comparison to lesser draft picks, and high production players are a scarce resource.

        • 9/22/2010
          Reply

          We disagree here. It really does depend on the player available and as I said it’s fairly easy to predict who will be available were within a few picks.

  6. Xavier Q
    9/20/2010
    Reply

    Austin,

    The issue doesn’t seem to be to get the best talent. If you have a #1 pick then 100% of TT5 guys are available, so it’s the best pick to have. The issue is value, since #1 picks get paid substantially more than #10 picks.

    There is somewhat of a performance bias, in that later picks don’t get playing time if they don’t perform, whereas higher picks get playing time no matter what. But that only comfirms that the lottery picks are playing poorly, while also eliminating that 30th pick that got 2 minutes of playing time and scored 2 points off an offensive rebound.

    Just looking at the numbers it is intuitive that GM’s do not know how to select talent. In 3 of the 7 sample years the #1 pick wasn’t in the top 12 performing players over their rookie contract. So if it is in essence simply chance whether you land a top talent or not, it would be preferable to select later in the round and pay that pick less money. And if you have a fairly accurate measure to predict which players are going to be productive, like PAWS48, then those later picks are even more valuable since the people picking ahead of you are likely to pass over at least one or two of the top players. Again, so you can pay the player less and get more value for your money.

    • 9/20/2010
      Reply

      Austin,
      You’ll note that I focused on per minute performance and took out players with small samples to even out the noise. I’m actually of the opinion that a lot of decent players are slipping through the cracks in the current system but that’s not the central issue. My conclusion is about market inefficiencies at identifying talent and how that drives TT5 player availability to the later part of the first.
      XQ,
      You nailed it .
      While that market inefficency exists, lower draft picks will always have more value given that there’s a huge drop-off in financial commitment and public expectation (my risk) versus no appreciable drop-off in potential value. If I can put a model together that has a better (say a 50% percent) chance of id’ing TT5 talent? Then the goal would be to identify undervalued talent and draft it as late as possible to maximize my possible value by mitigating my risk (salary and expectations) while possibly also getting value in return.

      Let’s give an example from real life. The Twolves supposedly coveted Evan Turner. If I’m the GM for the 76ers and KHAAAANNN calls me with an offer of the 4, 16 and 30 for the number 2, I cover up the phone to stop from laughing and take that to the bank (in my mind that’s a Turner for Cousins, Damion Jones and Whiteside trade). A 2 for 4 and 16 would also be a no brainer ( and here I could probably get the Twolves to throw in some sort of future pick or cash compensations).

      • StLreflections
        9/22/2010
        Reply

        The opportunity costs of the hype plus extra expense of a number 1 pick are negligible in comparison to the value of high performance. Rookie contracts are all vastly below market value for star players. As an NBA GM I’d happily trade an increased chance at a superstar in exchange for some bad publicity after a bust.

        Also, of course, limiting things to the 1st 4 years underestimates the value of some stars. e.g., Josh Howard is much less valuable than Bosh, LeBron, and Wade today. Even if you only have absolute control for the first (actually 5 years) of a contract, drafting a player has real advantages in terms of controlling the resource, both increasing the possibility of keeping that player around, and also giving you the ability to trade them for other valuable assets.

        I’m looking forward to seeing your method for picking top 5 guys/draft evaluation-if you can hit 50% on the late first round/second round breakouts we’ve seen over the last 15 years (that is, both correctly identifying who is undervalued, and not ending up with lots of false positives), that’s one thing-then I do think you can draft around that strategy, but if your odds don’t change things that much, then its not as effective. The raw numbers matter a lot in evaluating your method. Right now, its obvious that #1 draft picks are tremendously valuable, since they are so much more likely to provide high production in comparison to lesser draft picks, and high production players are a scarce resource.

        • 9/22/2010
          Reply

          We disagree here. It really does depend on the player available and as I said it’s fairly easy to predict who will be available were within a few picks.

  7. Evanz
    9/20/2010
    Reply

    Rookie contracts are incredibly cheap. The #1 pick gets roughly $4.5M. The difference between the #1 pick and the last pick of the first round is about $3M. In a typical free agent year, how many starters are available for $3M? Not that many. Taking my Warriors as an example, we just picked up Dorell Wright for around $3.5M. I think that’s an incredibly fortunate circumstance, though. Normally, it’s slim pickings at that level. I think I would still take my chances with a top 5 pick, rather than trading down, to save a couple million bucks.

    • 9/20/2010
      Reply

      Evanz,
      But the *perceived* value of the top 5 picks is so inflated that I could bilk someone for picks and players and still walk away from the draft with the player my model predicted would be good. The contract isn’t a huge issue but i’d rather spend that 3 million on a kick-ass stats department and stealing the Phoenix training staff.

      • Evanz
        9/20/2010
        Reply

        It seems to me that in the long run, your team is unlikely to get a true “franchise player” (Duncan, Shaq, LeBron, Howard, Durant, etc) by always trading down. If I were in that position, I don’t think I could trade down that kind of opportunity for a stats department. No offense, of course. 😉

        • 9/20/2010
          Reply

          I can name a few franchise players (>.250 WP48, >12 wins) that buck that trend: Barkley(5), Bird(6), Paul(4), Rodman(27), Marion(9), Stockton(16), Rondo(21),Boozer (34)) and that’s not everybody.

          • evanz
            9/20/2010
            Reply

            Barkley, Paul, and Bird are Top 6. That’s not really that low. Rodman, Marion, Stockton, Rondo, and Boozer are not franchise player. None of them could do what MJ or LeBron did. You’re way overselling those players.

      • Evanz
        9/20/2010
        Reply

        Another point that needs to be made is that in the 70’s, kids had to stay in college. That meant that players reached their potential much sooner upon coming into the NBA. I think the early entry and now this dumb one-year rule, are delaying the learning curve for incoming players. Going back again to a guy like Dorell Wright. It’s incredible that after 6 years in the league after being drafted out of high school, he’s still only 24 and only now getting a true shot at a significant role. The other side to this is that WP48 accounts for productivity, but not necessarily talent level. Some very rare specimens like LeBron or Duncan can be great right away, whereas a guy like Dwight Howard takes a few years. I think that skews the stats such that if you are looking at the first four years after entry, the very young top draft picks may have a disadvantage compared to some of the later round draft picks who stay 3 or 4 years and are ready to go when they get to the League.

        OTOH, I know you are interested in those first four years, because that is the rookie contract. However, when it comes to franchise players, if you land one, you’re going to do whatever it takes to keep those guys. Well, that is, if you’re not Charlotte or Cleveland.

        • 9/20/2010
          Reply

          Actually, Milwakee (Kareem), Memphis (Gasol), Minn (Garnett), Orlando (Shaq), LA (Shaq), Philadelphia (Wilt), Charlotte (Mourning), Toronto (Camby), Washington (Wallace), Barkley (Philly), Rodman (Detroit) the list goes on and on for Franchise guys leaving or being let go for lack of understanding of their value.

          • evanz
            9/20/2010
            Reply

            Notice most of the teams you just mentioned are small market and had cheap owners. LA let Shaq go because of the Kobe-Shaq feud, more than anything else. At least, that’s my take on what happened.

            You really consider Rodman a “franchise” player that should be mentioned in the same breath as MJ, Shaq, LeBron, or Duncan?

            • 9/20/2010
              Reply

              You mean 5 time NBA Champion, 7 time rebounding Champion and the player Phil Jackson called the greatest athlete he ever coached, arguably the best defender and definitely the best rebounder since the merger, the second best player on the best team ever and the guy who put together the most insane Wins Produced season ever, that Rodman? Yes I would.

          • evanz
            9/20/2010
            Reply

            Also Garnett, Kareem, Rodman and Barkley were let go well after their rookie contracts.

            • 9/20/2010
              Reply

              But the point was that you can go out and get a franchise player (particularly if you have a team of high value,low cost players).

              • evanz
                9/20/2010

                That may be true. But the teams who are able to get those players (LA or Boston, uh, now Miami) are typically not the ones at the top of the lottery (LAC, SAC, WAS, GSW). OKC changed its whole franchise by drafting Durant and Westbrook. The odds of OKC getting players of that quality in the free agent market would’ve been fairly minute.

              • Spider Jerusalem
                9/20/2010

                *ahem*

                Seattle drafted Durant and Westbrook. Let us not forget.

                To your main point, I think it brings to light the fact that not all draft classes are equal. In 2003, you’d be a fool to trade out of the top 3 spots (unless you’re Joe Dumars, then you’d be doing yourself a favor), knowing that Lebron, Wade and Bosh were all preeminent talents and all hyped to the degree that they wouldn’t be available later. However, in 2002, you’d be a fool to stay in spots 2-5 and overpay for marginal players.

                Arturo is dead on with the risk assessment in terms of hype, though. Regardless of how weak a class is, every single one is billed as being filled to the brim with superstars, when that often isn’t the case. And even when it is (like in 2005 and 2006), often the wrong guys will get tabbed as the cornerstones. There’s the assumption that if a guy is taken in the top 5, he has to be a franchise player. In those years (which seem pretty typical), I think Arturo’s strategy has no equal. In those rare cases where the hype meets up with talent (2003 with Lebron, 2004 with Howard, etc.) it’s a harder call to make.

              • evanz
                9/20/2010

                If the class is weak, it’s harder to trade down for anything of value, though. If the class is strong, then as you said, you don’t want to drop down.

                In either case, the assumption is that you would have a metric for picking the right player, as was said above. If you have that metric, then you are still better off staying at the top and picking the best player available. And if the class is so weak, that nobody of “franchise” quality is available at your pick (say 3,4,5), then what are you going to trade down for? Every GM knows when it’s a weak year. We all do.

                This past draft was a good example. It was a very weak draft after the Top 5. This would have been a great season to put to Arturo’s method. If you were the GM, and say had the #1 pick, do you trade down and not take Wall? or Turner? Johnson? Favors? Cousins? Which is the guy that you say you’re willing to pass up? I’d like to hear Arturo’s thoughts on that.

              • 9/20/2010

                evanz,
                I thought I’d been emphatic that I’d want Cousins in my wizards post. His PAWS40 was great, he’s polished, he’s big and I could get him underpriced with lowered expectations. I’ve mentioned some others but don’t worry the rookie model is coming (including the odds-on favorite to win rookie of the year).

  8. Evanz
    9/20/2010
    Reply

    Rookie contracts are incredibly cheap. The #1 pick gets roughly $4.5M. The difference between the #1 pick and the last pick of the first round is about $3M. In a typical free agent year, how many starters are available for $3M? Not that many. Taking my Warriors as an example, we just picked up Dorell Wright for around $3.5M. I think that’s an incredibly fortunate circumstance, though. Normally, it’s slim pickings at that level. I think I would still take my chances with a top 5 pick, rather than trading down, to save a couple million bucks.

    • 9/20/2010
      Reply

      Evanz,
      But the *perceived* value of the top 5 picks is so inflated that I could bilk someone for picks and players and still walk away from the draft with the player my model predicted would be good. The contract isn’t a huge issue but i’d rather spend that 3 million on a kick-ass stats department and stealing the Phoenix training staff.

      • Evanz
        9/20/2010
        Reply

        It seems to me that in the long run, your team is unlikely to get a true “franchise player” (Duncan, Shaq, LeBron, Howard, Durant, etc) by always trading down. If I were in that position, I don’t think I could trade down that kind of opportunity for a stats department. No offense, of course. 😉

        • 9/20/2010
          Reply

          I can name a few franchise players (>.250 WP48, >12 wins) that buck that trend: Barkley(5), Bird(6), Paul(4), Rodman(27), Marion(9), Stockton(16), Rondo(21),Boozer (34)) and that’s not everybody.

          • evanz
            9/20/2010
            Reply

            Barkley, Paul, and Bird are Top 6. That’s not really that low. Rodman, Marion, Stockton, Rondo, and Boozer are not franchise player. None of them could do what MJ or LeBron did. You’re way overselling those players.

      • Evanz
        9/20/2010
        Reply

        Another point that needs to be made is that in the 70’s, kids had to stay in college. That meant that players reached their potential much sooner upon coming into the NBA. I think the early entry and now this dumb one-year rule, are delaying the learning curve for incoming players. Going back again to a guy like Dorell Wright. It’s incredible that after 6 years in the league after being drafted out of high school, he’s still only 24 and only now getting a true shot at a significant role. The other side to this is that WP48 accounts for productivity, but not necessarily talent level. Some very rare specimens like LeBron or Duncan can be great right away, whereas a guy like Dwight Howard takes a few years. I think that skews the stats such that if you are looking at the first four years after entry, the very young top draft picks may have a disadvantage compared to some of the later round draft picks who stay 3 or 4 years and are ready to go when they get to the League.

        OTOH, I know you are interested in those first four years, because that is the rookie contract. However, when it comes to franchise players, if you land one, you’re going to do whatever it takes to keep those guys. Well, that is, if you’re not Charlotte or Cleveland.

        • 9/20/2010
          Reply

          Actually, Milwakee (Kareem), Memphis (Gasol), Minn (Garnett), Orlando (Shaq), LA (Shaq), Philadelphia (Wilt), Charlotte (Mourning), Toronto (Camby), Washington (Wallace), Barkley (Philly), Rodman (Detroit) the list goes on and on for Franchise guys leaving or being let go for lack of understanding of their value.

          • evanz
            9/20/2010
            Reply

            Notice most of the teams you just mentioned are small market and had cheap owners. LA let Shaq go because of the Kobe-Shaq feud, more than anything else. At least, that’s my take on what happened.

            You really consider Rodman a “franchise” player that should be mentioned in the same breath as MJ, Shaq, LeBron, or Duncan?

            • 9/20/2010
              Reply

              You mean 5 time NBA Champion, 7 time rebounding Champion and the player Phil Jackson called the greatest athlete he ever coached, arguably the best defender and definitely the best rebounder since the merger, the second best player on the best team ever and the guy who put together the most insane Wins Produced season ever, that Rodman? Yes I would.

          • evanz
            9/20/2010
            Reply

            Also Garnett, Kareem, Rodman and Barkley were let go well after their rookie contracts.

            • 9/20/2010
              Reply

              But the point was that you can go out and get a franchise player (particularly if you have a team of high value,low cost players).

              • evanz
                9/20/2010

                That may be true. But the teams who are able to get those players (LA or Boston, uh, now Miami) are typically not the ones at the top of the lottery (LAC, SAC, WAS, GSW). OKC changed its whole franchise by drafting Durant and Westbrook. The odds of OKC getting players of that quality in the free agent market would’ve been fairly minute.

              • Spider Jerusalem
                9/20/2010

                *ahem*

                Seattle drafted Durant and Westbrook. Let us not forget.

                To your main point, I think it brings to light the fact that not all draft classes are equal. In 2003, you’d be a fool to trade out of the top 3 spots (unless you’re Joe Dumars, then you’d be doing yourself a favor), knowing that Lebron, Wade and Bosh were all preeminent talents and all hyped to the degree that they wouldn’t be available later. However, in 2002, you’d be a fool to stay in spots 2-5 and overpay for marginal players.

                Arturo is dead on with the risk assessment in terms of hype, though. Regardless of how weak a class is, every single one is billed as being filled to the brim with superstars, when that often isn’t the case. And even when it is (like in 2005 and 2006), often the wrong guys will get tabbed as the cornerstones. There’s the assumption that if a guy is taken in the top 5, he has to be a franchise player. In those years (which seem pretty typical), I think Arturo’s strategy has no equal. In those rare cases where the hype meets up with talent (2003 with Lebron, 2004 with Howard, etc.) it’s a harder call to make.

              • evanz
                9/20/2010

                If the class is weak, it’s harder to trade down for anything of value, though. If the class is strong, then as you said, you don’t want to drop down.

                In either case, the assumption is that you would have a metric for picking the right player, as was said above. If you have that metric, then you are still better off staying at the top and picking the best player available. And if the class is so weak, that nobody of “franchise” quality is available at your pick (say 3,4,5), then what are you going to trade down for? Every GM knows when it’s a weak year. We all do.

                This past draft was a good example. It was a very weak draft after the Top 5. This would have been a great season to put to Arturo’s method. If you were the GM, and say had the #1 pick, do you trade down and not take Wall? or Turner? Johnson? Favors? Cousins? Which is the guy that you say you’re willing to pass up? I’d like to hear Arturo’s thoughts on that.

              • 9/20/2010

                evanz,
                I thought I’d been emphatic that I’d want Cousins in my wizards post. His PAWS40 was great, he’s polished, he’s big and I could get him underpriced with lowered expectations. I’ve mentioned some others but don’t worry the rookie model is coming (including the odds-on favorite to win rookie of the year).

  9. Tom Mandel
    9/21/2010
    Reply

    The first comment is pretty definitive. You are working with a selection set that *inevitably* skews the results by failing to account for draftees who don’t play the minutes.

    You might at some point turn to an analysis of guys who make it in the league despite not being drafted at all. This would provide a radical instance of the above. Given the low cost of these guys, one could argue that GMs should skip the draft.

    Another basic problem is that the map of relative success after 4 years cannot be fed back uncritically as an indicator of a better draft order. *Many* things happen in 4 years of life and basketball that could not have been predicted. Yet, random events occurring to a very high draft choice have a bigger effect on your conclusions than reandom events happening to guys lower down in the draft.

    How, for example, does the case of Len Bias affect your results for the year he was drafted? Compared to some guy drafted in the 20s that year who didn’t play. How does (#7 pick) Bobby Hurley’s accident affect your data compared to the fact that #26 pick Geert Hamminck didn’t play in the league?

    • 9/21/2010
      Reply

      Tom,
      If I ever do by some miracle wind up in an NBA front office, I’d hire you just to ask the tough questions :-) I have done the undrafted/draftee/free agent analysis in terms of building a team (I’ve just never posted it). Some teams (San Antonio for example) do a fantastice jo of finding undrafted talent and developing it. This, and the fact that all my data suggests that there are good players in college who do not get a chance in the pros is the reason that in the manifesto I would focus so much attention on my d-league teams to find these guys.

      I agree that there is noise in the data that skews the results somewhat (Bias, Hurley, good players not getting a shot) but you’ll note that I focused on per minute numbers of guys who played a significant amount of minutes and took a large data set. This minimizes the skew and the conclusions are definite enough do as for it not to matter.

  10. Tom Mandel
    9/21/2010
    Reply

    The first comment is pretty definitive. You are working with a selection set that *inevitably* skews the results by failing to account for draftees who don’t play the minutes.

    You might at some point turn to an analysis of guys who make it in the league despite not being drafted at all. This would provide a radical instance of the above. Given the low cost of these guys, one could argue that GMs should skip the draft.

    Another basic problem is that the map of relative success after 4 years cannot be fed back uncritically as an indicator of a better draft order. *Many* things happen in 4 years of life and basketball that could not have been predicted. Yet, random events occurring to a very high draft choice have a bigger effect on your conclusions than reandom events happening to guys lower down in the draft.

    How, for example, does the case of Len Bias affect your results for the year he was drafted? Compared to some guy drafted in the 20s that year who didn’t play. How does (#7 pick) Bobby Hurley’s accident affect your data compared to the fact that #26 pick Geert Hamminck didn’t play in the league?

    • 9/21/2010
      Reply

      Tom,
      If I ever do by some miracle wind up in an NBA front office, I’d hire you just to ask the tough questions :-) I have done the undrafted/draftee/free agent analysis in terms of building a team (I’ve just never posted it). Some teams (San Antonio for example) do a fantastice jo of finding undrafted talent and developing it. This, and the fact that all my data suggests that there are good players in college who do not get a chance in the pros is the reason that in the manifesto I would focus so much attention on my d-league teams to find these guys.

      I agree that there is noise in the data that skews the results somewhat (Bias, Hurley, good players not getting a shot) but you’ll note that I focused on per minute numbers of guys who played a significant amount of minutes and took a large data set. This minimizes the skew and the conclusions are definite enough do as for it not to matter.

  11. Tom Mandel
    9/21/2010
    Reply

    This is your key point: “the *perceived* value of the top 5 picks is so inflated that I could bilk someone for picks and players” by trading down and get much more for my money.

    That is certainly true — especially in retrospect, where after the draft we can trade for the guys we now know would be available at the picks we trade for. And even more especially in later retrospect, where we can say we would have traded for those guys who turned out well.

    When it’s prospective players, the issue is not so clear. The Len Bias case is extreme but it suffices to make the point. By definition you cannot correct for future random effects. Your dataset has corrected for them retrospectively. It’s a problem.

    If you have a model that lets you predict that an individual player has a higher chance of future success than his likely draft position, then your model says “trade to get the pick at that position”. It doesn’t necessarily say trade *down* btw. You may have to trade up.

    Think about this anomaly: pick #2 seems over time to have been less valuable than pick #3. Does this mean that if you have pick #2, you should swap it straight up for pick #3? This is what your analysis argues, but of course it’s absurd.

    • 9/21/2010
      Reply

      Tom,
      The trick is to create a consistent draft strategy that has a decent hit rate (not perfect but decent) and stick to it. It’s easy enough to predict were teams are going to go and do your trades accordingly. It really isn’t about the position as much as maximizing the value you get on draft day. If you really like a player and you need to move up to get him, you do as long as the value/risk equation still works out in your favor (can I give somebody an overvalued scorer perhaps?).

      I think the number 2 pick is simply a reflection of the poor ability to id talent. The number 1 pick from 2000-2007 got you top 2 talent 3 times, top 12 once, and wiffed 3 times. Number two hit the top once and top 5 another time (we can probably call 2007 as well). Picking the a good guy as number one is hard. Picking two guys as 1-2, incredibly so.

  12. Tom Mandel
    9/21/2010
    Reply

    This is your key point: “the *perceived* value of the top 5 picks is so inflated that I could bilk someone for picks and players” by trading down and get much more for my money.

    That is certainly true — especially in retrospect, where after the draft we can trade for the guys we now know would be available at the picks we trade for. And even more especially in later retrospect, where we can say we would have traded for those guys who turned out well.

    When it’s prospective players, the issue is not so clear. The Len Bias case is extreme but it suffices to make the point. By definition you cannot correct for future random effects. Your dataset has corrected for them retrospectively. It’s a problem.

    If you have a model that lets you predict that an individual player has a higher chance of future success than his likely draft position, then your model says “trade to get the pick at that position”. It doesn’t necessarily say trade *down* btw. You may have to trade up.

    Think about this anomaly: pick #2 seems over time to have been less valuable than pick #3. Does this mean that if you have pick #2, you should swap it straight up for pick #3? This is what your analysis argues, but of course it’s absurd.

    • 9/21/2010
      Reply

      Tom,
      The trick is to create a consistent draft strategy that has a decent hit rate (not perfect but decent) and stick to it. It’s easy enough to predict were teams are going to go and do your trades accordingly. It really isn’t about the position as much as maximizing the value you get on draft day. If you really like a player and you need to move up to get him, you do as long as the value/risk equation still works out in your favor (can I give somebody an overvalued scorer perhaps?).

      I think the number 2 pick is simply a reflection of the poor ability to id talent. The number 1 pick from 2000-2007 got you top 2 talent 3 times, top 12 once, and wiffed 3 times. Number two hit the top once and top 5 another time (we can probably call 2007 as well). Picking the a good guy as number one is hard. Picking two guys as 1-2, incredibly so.

  13. Man of Steele
    9/21/2010
    Reply

    Nice article. I wonder how you might develop a metric to help you wade through the pre-draft hype. As mentioned, the hyperbole around every draft class is overwhelming, but it is manifestly the case that not all draft classes are strong. In the rare draft classes that truly are strong at the top, however, it would be favorable to trade up. For example, in the James/Wade/Bosh draft, it would have been nice to have a top 3 pick (since the best players were available only at the top of the draft). In other years, though, the gap between the actual and perceived values of the top picks is wider.

    Or look at it this way: someone commented that, besides later draft picks, there exists another class of undervalued players. Players like LeBron James or Chris Paul, who are worth much more than a maximum contract, are actually great values as well. Getting a guy like this for a 4-yr rookie contract, and then probably having a good chance of resigning him for a good price afterward, is obviously the most desirable conerstone for a team. Such a pipe dream would only be realists, though, if you were able to distinguish clealry between relative strengths of various draft calsses.

    • 9/21/2010
      Reply

      MOS,
      There’s always the option of trading for future picks from bad teams (which we can call using our model). The 2011 picks of the 76ers,Wiz, Twolves,Pacers and Raptors would all be very intriguing. You only need to win the lottery once every four years.

  14. Man of Steele
    9/21/2010
    Reply

    Nice article. I wonder how you might develop a metric to help you wade through the pre-draft hype. As mentioned, the hyperbole around every draft class is overwhelming, but it is manifestly the case that not all draft classes are strong. In the rare draft classes that truly are strong at the top, however, it would be favorable to trade up. For example, in the James/Wade/Bosh draft, it would have been nice to have a top 3 pick (since the best players were available only at the top of the draft). In other years, though, the gap between the actual and perceived values of the top picks is wider.

    Or look at it this way: someone commented that, besides later draft picks, there exists another class of undervalued players. Players like LeBron James or Chris Paul, who are worth much more than a maximum contract, are actually great values as well. Getting a guy like this for a 4-yr rookie contract, and then probably having a good chance of resigning him for a good price afterward, is obviously the most desirable conerstone for a team. Such a pipe dream would only be realists, though, if you were able to distinguish clealry between relative strengths of various draft calsses.

    • 9/21/2010
      Reply

      MOS,
      There’s always the option of trading for future picks from bad teams (which we can call using our model). The 2011 picks of the 76ers,Wiz, Twolves,Pacers and Raptors would all be very intriguing. You only need to win the lottery once every four years.

  15. neal frazier
    9/22/2010
    Reply

    So if I am an old school gm who is basically throwing darts, I want the high pick, but if you have a good predictive model you want the lower picks – sounds like an opportunity for a win win trade – as an incompetent GM I can improve my odds by trading up at the same time that you (competent GM) improve your odds by trading down – comparative advantage at work :)

    • 9/23/2010
      Reply

      Yep, and the old school gm might throw some undervalued assets in the trade as well.

  16. neal frazier
    9/22/2010
    Reply

    So if I am an old school gm who is basically throwing darts, I want the high pick, but if you have a good predictive model you want the lower picks – sounds like an opportunity for a win win trade – as an incompetent GM I can improve my odds by trading up at the same time that you (competent GM) improve your odds by trading down – comparative advantage at work :)

    • 9/23/2010
      Reply

      Yep, and the old school gm might throw some undervalued assets in the trade as well.

  17. […] point: most NBA teams are bad at drafting. Players like Leonard and Faried frequently last past the middle of the first round. Charlotte, Cleveland, Minnesota, Utah, and Washington all had the opportunity to acquire Leonard […]

  18. […] point: most NBA teams are bad at drafting. Players like Leonard and Faried frequently last past the middle of the first round. Charlotte, Cleveland, Minnesota, Utah, and Washington all had the opportunity to acquire Leonard […]

Leave a Reply

Your email address will not be published. Required fields are marked *